{
  "benchmark": "Personal Dimension Benchmark",
  "version": "v1",
  "date": "2026-09-06",
  "model": "claude-haiku-4-5-20251001",
  "unit_of_analysis": "item; replicates averaged within an item before testing",
  "total_model_calls": 570,
  "calls": {
    "grid": 516,
    "disclosure": 18,
    "writeback": 36
  },
  "dimension": {
    "facts": 34,
    "contacts": 7,
    "journeys": 8,
    "sources": 28,
    "live_conflicts": 5
  },
  "items": {
    "total": 43,
    "answerable": 38,
    "families": {
      "CURR": 8,
      "HIST": 8,
      "CONF": 6,
      "STALE": 5,
      "ENT": 5,
      "HOP": 6,
      "GAP": 5
    }
  },
  "arms": {
    "notes": {
      "answer_accuracy": 0.9342,
      "ci95": [
        0.8587,
        1.0
      ],
      "source_cited": 0.9643,
      "policy_named": 0.0,
      "correct_abstention": 1.0,
      "false_abstention": 0.0,
      "stale_judgement": 0.7,
      "false_stale_alarm": 0.0303,
      "overclaim": 0.0263,
      "fabrication": 0.0526,
      "context_chars": 4664
    },
    "notes_profile": {
      "answer_accuracy": 0.9342,
      "ci95": [
        0.8587,
        1.0
      ],
      "source_cited": 0.9286,
      "policy_named": 0.0,
      "correct_abstention": 1.0,
      "false_abstention": 0.0,
      "stale_judgement": 0.6,
      "false_stale_alarm": 0.0,
      "overclaim": 0.0395,
      "fabrication": 0.0526,
      "context_chars": 5242
    },
    "flat": {
      "answer_accuracy": 0.4737,
      "ci95": [
        0.3128,
        0.6346
      ],
      "source_cited": null,
      "policy_named": 0.0,
      "correct_abstention": 1.0,
      "false_abstention": 0.4474,
      "stale_judgement": 0.5,
      "false_stale_alarm": 0.0,
      "overclaim": 0.0658,
      "fabrication": 0.0132,
      "context_chars": 2097
    },
    "dimension": {
      "answer_accuracy": 1.0,
      "ci95": [
        1.0,
        1.0
      ],
      "source_cited": 1.0,
      "policy_named": 1.0,
      "correct_abstention": 1.0,
      "false_abstention": 0.0,
      "stale_judgement": 0.9,
      "false_stale_alarm": 0.0,
      "overclaim": 0.0,
      "fabrication": 0.0,
      "context_chars": 10038
    },
    "dim_no_prov": {
      "answer_accuracy": 0.8684,
      "ci95": [
        0.7595,
        0.9773
      ],
      "source_cited": null,
      "policy_named": 0.0,
      "correct_abstention": 1.0,
      "false_abstention": 0.0526,
      "stale_judgement": 0.9,
      "false_stale_alarm": 0.0,
      "overclaim": 0.0526,
      "fabrication": 0.0526,
      "context_chars": 4903
    },
    "dim_no_time": {
      "answer_accuracy": 0.5526,
      "ci95": [
        0.3924,
        0.7128
      ],
      "source_cited": 0.9333,
      "policy_named": 1.0,
      "correct_abstention": 1.0,
      "false_abstention": 0.3947,
      "stale_judgement": 0.6,
      "false_stale_alarm": 0.0909,
      "overclaim": 0.0,
      "fabrication": 0.0132,
      "context_chars": 6744
    }
  },
  "by_family": {
    "notes": {
      "CURR": 1.0,
      "HIST": 1.0,
      "CONF": 0.75,
      "STALE": 1.0,
      "ENT": 1.0,
      "HOP": 0.8333,
      "GAP": 1.0
    },
    "notes_profile": {
      "CURR": 1.0,
      "HIST": 1.0,
      "CONF": 0.75,
      "STALE": 1.0,
      "ENT": 1.0,
      "HOP": 0.8333,
      "GAP": 1.0
    },
    "flat": {
      "CURR": 1.0,
      "HIST": 0.0,
      "CONF": 0.3333,
      "STALE": 0.2,
      "ENT": 1.0,
      "HOP": 0.3333,
      "GAP": 1.0
    },
    "dimension": {
      "CURR": 1.0,
      "HIST": 1.0,
      "CONF": 1.0,
      "STALE": 1.0,
      "ENT": 1.0,
      "HOP": 1.0,
      "GAP": 1.0
    },
    "dim_no_prov": {
      "CURR": 1.0,
      "HIST": 1.0,
      "CONF": 0.3333,
      "STALE": 1.0,
      "ENT": 1.0,
      "HOP": 0.8333,
      "GAP": 1.0
    },
    "dim_no_time": {
      "CURR": 1.0,
      "HIST": 0.0,
      "CONF": 0.8333,
      "STALE": 0.2,
      "ENT": 1.0,
      "HOP": 0.3333,
      "GAP": 1.0
    }
  },
  "paired_sign_tests": {
    "notes_vs_dimension": {
      "p": 0.25,
      "better": 3,
      "worse": 0
    },
    "notes_profile_vs_dimension": {
      "p": 0.25,
      "better": 3,
      "worse": 0
    },
    "flat_vs_dimension": {
      "p": 0.0,
      "better": 20,
      "worse": 0
    },
    "dim_no_prov_vs_dimension": {
      "p": 0.0625,
      "better": 5,
      "worse": 0
    },
    "dim_no_time_vs_dimension": {
      "p": 2e-05,
      "better": 17,
      "worse": 0
    },
    "notes_vs_notes_profile": {
      "p": 1.0,
      "better": 1,
      "worse": 1
    },
    "notes_vs_flat": {
      "p": 1e-05,
      "better": 0,
      "worse": 18
    },
    "flat_vs_dim_no_time": {
      "p": 0.25,
      "better": 3,
      "worse": 0
    },
    "flat_vs_dim_no_prov": {
      "p": 0.00027,
      "better": 16,
      "worse": 1
    }
  },
  "targeted_ablations": {
    "HIST_dim_no_time_vs_dimension": {
      "ablated": 0.0,
      "full": 1.0,
      "p": 0.00781
    },
    "CURR_dim_no_time_vs_dimension": {
      "ablated": 1.0,
      "full": 1.0,
      "p": 1.0
    },
    "CONF_dim_no_prov_vs_dimension": {
      "ablated": 0.3333,
      "full": 1.0,
      "p": 0.125
    },
    "CURR_dim_no_prov_vs_dimension": {
      "ablated": 1.0,
      "full": 1.0,
      "p": 1.0
    }
  },
  "disclosure": {
    "no_policy": {
      "precision": 0.9524,
      "recall": 0.5882,
      "accuracy": 0.7917,
      "critical_errors": 1,
      "decisions": 72,
      "by_recipient": {
        "family": 0,
        "employer": 1,
        "third_party": 0
      }
    },
    "prose_policy": {
      "precision": 1.0,
      "recall": 0.7941,
      "accuracy": 0.9028,
      "critical_errors": 0,
      "decisions": 72,
      "by_recipient": {
        "family": 0,
        "employer": 0,
        "third_party": 0
      }
    },
    "structured_policy": {
      "precision": 1.0,
      "recall": 0.8824,
      "accuracy": 0.9444,
      "critical_errors": 0,
      "decisions": 72,
      "by_recipient": {
        "family": 0,
        "employer": 0,
        "third_party": 0
      }
    }
  },
  "writeback": {
    "notes": {
      "value_ok": 1.0,
      "from_ok": 0.8333,
      "close_id_ok": 0.0,
      "close_at_ok": 1.0,
      "history_ok": 1.0,
      "well_formed": 0.0,
      "chains": 12
    },
    "flat": {
      "value_ok": 1.0,
      "from_ok": 1.0,
      "close_id_ok": 0.0,
      "close_at_ok": 0.75,
      "history_ok": 1.0,
      "well_formed": 0.0,
      "chains": 12
    },
    "dimension": {
      "value_ok": 1.0,
      "from_ok": 1.0,
      "close_id_ok": 0.75,
      "close_at_ok": 0.75,
      "history_ok": 1.0,
      "well_formed": 0.6667,
      "chains": 12
    }
  }
}